Playbook · Content · Example output
Example output: one guide, three drafts¶
This is what the repurpose playbook produces. One source goes in, channel-ready drafts come out, and nothing appears in any draft that isn't in the source.
The inputs
Source: this site's own guide, Why AI makes things up. Formats asked for: newsletter blurb, LinkedIn post, X thread. Audience: non-technical people using AI at work. Rules: plain words, no hype, no em-dashes. Everything below the line is the playbook's actual output for those answers.
Draft 1: newsletter blurb¶
The most dangerous AI answer is the confident one.
Sometimes an AI states something false with total confidence: a fact, a quote, a source that doesn't exist. It isn't lying and it isn't a glitch. The tool writes by predicting what text should come next, and it has no built-in step that asks "is this actually true?"
The fix is a checking habit, not a setting. Click the sources behind anything you'll act on. Check names, numbers, and dates against the original. Ask the same important question twice in fresh chats; two different confident answers means at least one is made up.
And skip the one that feels natural: asking "are you sure?" doesn't work. It will apologize and produce a new confident answer, which may also be wrong.
Draft 2: LinkedIn post¶
AI doesn't know when it's wrong. That's not a bug you can turn off.
These tools write by predicting what text should come next. They're excellent at producing text that looks right, and they have no internal step that checks whether it is right. When the AI doesn't know something, it fills the gap with something plausible, delivered in the same confident tone as a correct answer.
Where it bites hardest: citations, numbers, quotes, and niche topics with little written about them.
Three habits that catch it:
- Click the sources behind anything you'll act on. A link you can't open isn't a source.
- Ask the same important question a second time, fresh. Two different confident answers means at least one is invented.
- Say up front: "tell me what you don't know, never invent a source."
Confidence and correctness are unrelated. Treat every AI answer as a draft until the parts you'll act on are checked.
Draft 3: X thread¶
1/ An AI will state something false with total confidence. A fact, a quote, a source that doesn't exist. People call it hallucination. It's not lying, and there's no setting that turns it off.
2/ Why it happens: the AI writes by predicting what text should come next. It's very good at "looks right." It has no built-in step that asks "is this true?"
3/ So when it doesn't know, it doesn't feel the gap the way you would. It fills the gap with something plausible. Plausible is the problem: wrong arrives in the same tone as right.
4/ Where it bites hardest: invented citations, close-but-wrong numbers, quotes people never said, and niche topics where little material exists.
5/ What doesn't work: asking "are you sure?" It apologizes and produces a new confident answer. Which may also be wrong.
6/ What works: click every source you'll act on. Check names, numbers, dates. Ask the important question twice in fresh chats and compare.
7/ Treat every AI answer as a draft until the parts you'll act on are checked. Full guide, free, no signup: jordysupport.com
Why this output is trustworthy¶
Every claim in all three drafts traces back to the source guide. The playbook's rules forbid new facts and invented quotes, and it drafts one format at a time so you can react before it moves on. The Before you post anything checklist runs at the end.